News broke this week that the National Security Agency (NSA) is using Anthropic’s Mythos AI model for offensive cyber operations, likely against China and Iran. It makes a lot of sense, given the reported power of Mythos to find and exploit vulnerabilities. But it’s just the latest sign that Defense Secretary Pete Hegseth might be losing his war against Anthropic.
Anthropic has installed about half a dozen engineers at NSA, according to the Financial Times, though the newspaper reports it’s unclear if Anthropic staff are actively assisting in operations against adversarial countries. At the very least, they’re customizing the AI models for specific applications, according to the FT.
Back in March, the Pentagon labeled Anthropic a supply chain risk because the company refused to modify the guardrails on Claude. The Defense Department wanted to use Claude without any restrictions, but Anthropic wouldn’t budge on two issues: use of its AI for autonomous weapons and mass surveillance of Americans.
Anthropic CEO Dario Amodei met with Hegseth on Feb. 24, but they couldn’t come to an agreement and the Defense Secretary launched a crusade against the company to put it in its place. Anthropic wouldn’t budge, so President Donald Trump gave his blessing to blacklist the company, setting in motion a process to purge Claude from government systems. But it doesn’t seem like Hegseth’s war, waged in conjunction with Emil Michael, the Under Secretary of Defense for Research and Engineering, is going so well.
The argument that Hegseth and his underlings were trying to make, that Anthropic was a unique threat to national security, never made any sense. The company has been more than willing to work with the U.S. military and its allies, as has been demonstrated again and again.
Amodei visited in the White House in mid-April and Reuters has a new report Friday that the dispute between Anthropic and the Trump regime has been “showing signs of easing,” especially as the company prepares to go public. That’s important for Anthropic as it makes an initial public offering more safe for investors if the AI company has a good relationship with the government. But it also just makes sense for the military and intelligence communities, because they want access to the most advanced AI in the world.
Amodei was originally invited to the White House on May 21 for Trump’s planned signing of an executive order on AI, but that was derailed after the White House decided it didn’t like some of the provisions. But it will be interesting to watch the jockeying for power among the AI giants as they curry favor with Trump. OpenAI’s Sam Altman and xAI’s Elon Musk have their own feud going, and it can be tough to become the favorite oligarch when you have so many sociopathic billionaires in line for their government money.
NOTUS published a report Thursday that Altman has been floating the idea of the government taking a stake in OpenAI with Trump ever since the start of the president’s second term. The stake might be used for things “such as distributing a dividend payment to all American households,” according to NOTUS, an idea that’s become popular among AI enthusiasts who believe the tech will displace millions of workers and cause massive unemployment.
The U.S. government has already taken a stake in at least ten companies, according to the Wall Street Journal, in a highly unusual move typically reserved for extraordinary circumstances. The U.S. has a 10% stake in Intel and a golden share in U.S. Steel. But if the government takes a stake in OpenAI it sounds like that’s being pitched as something that could be used as a universal basic income or some other nominal form of financial benefit to Americans. The American people aren’t getting direct payments from Intel over that investment.
How wise would it be to tie the government’s fortune to huge AI companies? That remains to be seen, especially as it potential sets up incentives to benefit private companies over others and intertwine the economic stability of the government with a specific form of tech. It might seem smart to have the government invested in the latest and greatest, but tech is constantly changing.
Imagine if the government had invested heavily in VR headsets in the 2010s and bought Oculus, for example. The metaverse was pitched as the inevitable future, so much so that Facebook changed the name of its parent company to Meta. But that bet hasn’t panned out in many ways and it’s a good thing the U.S. taxpayer isn’t on the hook for what was always a big gamble.
Anthropic didn’t immediately respond to questions emailed Friday. Gizmodo will update this article if we hear back.
